Scholarlink

<p>Brain tumor segmentation is a critical task in medical image analysis, yet existing methods often struggle with accurately capturing complex tumor morphologies and delineating boundaries with high precision. Traditional approaches suffer from spatial inconsistencies and fail to effectively integrate multimodal MRI data. To address these challenges, this study introduces a Hybrid Attention-Enhanced U-Net, integrating spatial and channel recalibration mechanisms to improve feature representation and segmentation accuracy. The proposed model dynamically focuses on tumor-relevant regions while suppressing background noise, resulting in enhanced segmentation precision. Evaluated on the BraTS 2019, 2020, and 2023 datasets, the Hybrid Attention-Enhanced U-Net consistently achieved higher Dice and IoU scores compared to baseline models, demonstrating improved segmentation accuracy, particularly in capturing fine-grained tumor structures. Qualitative analysis further highlights the model’s ability to produce more accurate boundary delineations and reduced segmentation errors. This work represents a promising advancement in brain tumor segmentation, contributing to robust tumor detection and enhanced generalization in automated medical image analysis.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid attention enhanced U Net with anatomical gating for brain tumor segmentation using MRI images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月25日

查看全文
  • Abhishek Jadhav,
  • Akhtar Rasool,
  • Manasi Gyanchandani

摘要

Brain tumor segmentation is a critical task in medical image analysis, yet existing methods often struggle with accurately capturing complex tumor morphologies and delineating boundaries with high precision. Traditional approaches suffer from spatial inconsistencies and fail to effectively integrate multimodal MRI data. To address these challenges, this study introduces a Hybrid Attention-Enhanced U-Net, integrating spatial and channel recalibration mechanisms to improve feature representation and segmentation accuracy. The proposed model dynamically focuses on tumor-relevant regions while suppressing background noise, resulting in enhanced segmentation precision. Evaluated on the BraTS 2019, 2020, and 2023 datasets, the Hybrid Attention-Enhanced U-Net consistently achieved higher Dice and IoU scores compared to baseline models, demonstrating improved segmentation accuracy, particularly in capturing fine-grained tumor structures. Qualitative analysis further highlights the model’s ability to produce more accurate boundary delineations and reduced segmentation errors. This work represents a promising advancement in brain tumor segmentation, contributing to robust tumor detection and enhanced generalization in automated medical image analysis.